Pharma Algorithms

1

Pharma Algorithms
www.ap-algorithms.com
Classification Analysis (C-SAR) in Predicting
HUMAN INTESTINAL ABSORPTION
D. Zmuidinavicius, R. Didziapetris, P. Japertas,
A. Avdeef, A. Petrauskas
2
INTRODUCTION
Pharma Algorithms
1. C-SAR vs. QSAR
2. HIA Analysis
3. Using PAMPA
4. Building HTS Filters
3
1. What is C-SAR?
Pharma Algorithms
Based on recursive partitioning Groups compounds
into classes Aims at differentiating biol.
mechansisms Leads to the knowledge-based
predictions
?
?
?
?
4
Recursive Partitioning
Pharma Algorithms
A stepwise procedure One descriptor at a
time Finds the best cut-off value Finds the
best descriptor Repeats analysis for new classes
of compounds
?
?
?
?
?
(Example analysis of P-gp substrate
specificity)
5
Descriptors
Pharma Algorithms
Phys-Chem
Hansch-Leo
? Log P, Solubility, ? pKa, Ion form fractions ?
Abrahams solvation ? Lipinskis NHA, NHD ?
Ertls tPSA ? Vebers NRot-Bond ? MW, MolVol
? ? 103 descriptors
Functional groups interactions
Augmented Scaffolds
? ? 105 descriptors
Klopman-type biophores
Not used for HIA, used for P-gp
6
Advantages over QSAR
Pharma Algorithms
? Analyzes multiple mechanisms ? Uses all kinds
of descriptors ? Unlimited No of descriptors ?
Tolerates high data variability ? Focus -
Outlier analysis ? Replaces data mining
7
2. HIA Analysis
Pharma Algorithms
? gt 1,000 HIA values compiled ? gt 100 literature
sources used ? Abrahams compilation 270 HIAs ?
Therapeutic Drugs c.a. 600 HIAs ? Two-class
model (HIA10-15) ? Active transport 48
compounds
8
Stepwise Analysis
Pharma Algorithms
1. A series of C-SAR analyses 2. gtNlt, Nlt,
gtCPO(OH)22 do not absorb at any MW 3.
The remaining compounds
? Paracellular tr. (MW lt 255) ? Non-restricted
(MW lt 580) ? Restricted diff. (MW gt 580)
9
Examples of RP Trees
Pharma Algorithms
MW 255 - 580
MW gt 580
tPSA gt 154
tPSA gt 291
LogP gt 0
? gt 1.3
LogP gt 0
10
Open-Concept Filter
Pharma Algorithms
11
Permeability Prediction
Pharma Algorithms
Absorbable compounds
1. Driven by generic properties 2. Lipinskis
and Vebers rules
Cut-offs exceeded by 100 Diffusion not
differentiated Data mining, not C-SAR
?
?
?
12
HIA Prediction An Art of Exceptions
Pharma Algorithms
Analysis of Deviations
1. gtCPO(OH)22 just a beginning 2. Further
development
Substructure-specific rules (active tr., efflux,
P450) Exp. SW, PApp descriptors Use in F
predictions
?
?
?
13
3. Using PAMPA Data
Pharma Algorithms
14
C-SAR for PAMPA
Pharma Algorithms
? gt 0.65
tPSA gt 31
No charge
DATA PApp (cm/s, x106), pH 7.4, N
88 SOURCE Zhu et al, Eur J Med Chem, 2002, 37,
399-407
15
Binary Pe, N 316
Pharma Algorithms
Pe (cm/s, x106), cut-off Pe 10, pH 7.4
Compounds with no acidic groups, bearing
positive charge, tPSA lt 66, will permeate
16
C-SAR for F
Pharma Algorithms
Binary model F 30 N 812
AB/HIA
Zwitterions AT?
AB/Solubility
AB/LogP
17
Towards Mechanistic Descriptors
Pharma Algorithms
Replace theoretical permeability models with
exp.-based filters
18
In Silico ADME Model
Pharma Algorithms
Existing Filters
19
4. Building HTS Filters
Pharma Algorithms
? C-SAR/QSAR Builder ? Algorithm Builder ?
ADME/Tox Filters
? Booth ? Posters ? Articles